近年國內外對於建立土地使用變遷估計模型及探討驅動因子的相關研究眾多，考量的重點與採用工具及方法各異。本研究以二元羅吉斯特迴歸（BLR）、多元羅吉斯特迴歸（MLR）、線性迴歸（LR）等三種數理模型，建構臺北縣新店地區土地使用變遷估計模型，研究進一步將新店劃分為五個次區，探討不同估計模型，是否會因區域特性、都市發展管制強度與研究範圍之尺度等因素影響估計準確度。本研究成果發現BLR估計模型之「總正確估計涵蓋率」高於MLR再高於LR，亦即，估計第二期網格有無建物比估計建物面積的準確度高。除檢視模型的相關係數和「總正確估計涵蓋率」外，尚訂定五個建築用地指標，相互檢驗及比較模型的估計能力。結果發現（1）不同地區的確需要採用不同的數理模型；（2）建物用地變化率越大的地區，估計模型之總正確估計涵蓋率及R2值會越低；（3）估計全部建物用地量比估計建物用地產生變化之地區容易，（4）估計範圍擴大，估計建物用地變化的準確度會降低，表示尺度的確會影響模型估計能力；（5）BLR較LR模型適用於在建物用地變化比率較高的地方。

Many researches have been carried out to establish land use and land cover change (LUCC) model and derive variables that influence LUCC processes. The models developed were based on different considerations and numerical methods. This research employed binary logistic regression (BLR), multinomial logistic regression (MLR) and linear regression (LR) to predict the LUCC processes in Sindian city, Taipei County. The experimental area was sub-divided into five sub-areas for the assessment, and investigated whether the prediction results from the three methods would be affected by land use characteristics, intensity of land control as well as the size of model domain.
This research suggests that the accuracy of the prediction of total built-up area is highest for BLR, followed by MLR and LR. This study also establishes five indicators to evaluate and compare the accuracy of the predictions made by MLR, BLR and LR. It suggests that (1) different numerical methods should be used in different sub-areas; (2) the accuracy of the prediction of total built-up area and R2 value decreases as “built-up area change rate” of an area increases; (3) the three methods are comparatively better in predicting the total built-up area than the location of built-up area; (4) the prediction accuracy for total built-up area decreases as the size of model domain increases; (5) BLR appears to be more suitable to be used in the area with high “built-up area change rate” than LR does.